THE BOOK--Playing The Percentages In Baseball

Interesting article.

A few other factors immediately come to mind:

- players by team: many of the games in the low 60’s are going to come from Seattle, San Francisco and Oakland, the three coldest ballparks on average.  Hardly any will come from Florida, Atlanta or Texas, and none at all from the parks with retractable roofs (aside from Seattle which isn’t a complete enclosure, and Milwaukee which is covered but not heated)

- players by month: April is when the Spring Training phenoms start pitching to major leaguers full time, and there will be more marginal talent pitching in April than in July (my supposition, but I expect it to be proven true)

- potential small sample size on <50F games.  Same with very high temps, which also are going to only happen in certain places.

We’ve had a couple of similar-themed articles on the subject, which you can find here:

Wind:
http://www.insidethebook.com/ee/index.php/site/article/wind_patterns_affecting_stats/

Temperature:
http://www.insidethebook.com/ee/index.php/site/article/weather_park_factors/

(Note to Constancio: your image files have been moved, as the article shows a broken link.)

Greg, I know you’ve done similar work.  Did I see it in last year’s THT annual?  Or elsewhere?

You should expect to see smaller movement on pitches with warmer weather, as Jonathan shows.  Air density is inversely proportional to temperature, and the range of change in movement (about 10%) is also roughly consistent with what should be expected for the range of temperatures that baseball is played under.  Humidity should have a much smaller effect (1%).

One should also expect fly balls to travel a bit farther in warmer weather and for pitches to make it to the plate a bit quicker(1% or less) in warmer weather due to the air density argument.  However, in the pitchf/x data, there should be no change in the initial velocity of pitches as that has very little to do with air density, but there should be a small change upward in the speed of pitches as they cross the plate.

In Robert K Adair’s “The Physics of Baseball” he describes some of the effects of changing temperature. of particular interest are the measurements of temperature on the COR of the ball. A warmer baseball is livelier than a cold one. With the exception of very chilly weather, we do see an increase in HR rates with warmer temperatures. I suspect measurements of BABIP would also show an increase with temp. The spike at the coldest temperatures probably indicates the difficulty pitchers have trying to grip a ball and throw quality pitches with numb fingers.

Steve, check the links in post 2.

I have done some temperature studies over the years (not pitch movement of course) and there is no doubt that to draw any cogent and credible conclusions, you have to control for the players and perhaps even the month (even if you control for the players, if you don’t control for the month, you run the risk of mixing up effects of temperature with the effects of either early season rust or late season fatigue). Plus you also need to control for day/night as day games are quite a bit warmer but you also get the effect of the sun, lights, etc.  For example, according to my research, even after controlling for temperature, it appears that it is “easier” to hit during the day (increased offensive performance) than at night.

I am not sure why he discusses about the month by month graph the way he does.  He says that the fact that the numbers do not increase a lot as the season goes on indicates that it is temperature and not “shedding of rust” that drives the first graph (temp v. performance).  Well, as the season goes on, the temperature goes up, so you would EXPECT for the off performance to go up as the season goes on.  If anything, this is evidence that the PITCHERS are shedding rust as the season goes on.  More likely though, there is a problem with the pool of players.  What happens is that players get injured, young players are brought up in August and September, dropped after April, etc.  Plus, there is actually evidence that pitchers but not batters have a “shedding of rust” as the season goes on.

Basically, you can draw NO conclusions from these offense performance graphs about the effects of temperature without doing at least some of these controls.  Of course we already KNOW that increased temperature increases offensive performance, but without doing the controls and adjustments, we have NO idea of the correct nuances, spikes, magnitude or causes of the effects, etc.

Plus, I can see how curve balls with less movement would be a benefit to the hitters, but a fastball with less rise is not necessarily a detriment to the pitchers.  I think that the pitch f/x researchers have suggested that it is the movement of pitches (at least for fastballs) versus the expectation of batters that is the issue in terms of advantage.

This author says that HR rates go UP in very cold weather.  There could certainly be a point at which it is hard to grip the ball and that the hitter starts to get an advantage, but I would think that the lack of distance on the fly balls and the cold core would be more significant factors.  I think that sample size is an issue here since he is only working with one year of data.  I don’t recall in my research (I use many years of data) seeing any spike in HR rate at very cold temperatures, but I could be wrong.

As Tango said, this is OK as a first pass ONLY.

I wouldn’t mind having Greg chiming in again.

I don’t have all of my stuff at hand, unfortunately.  I know I’ve posted a chart showing the impact of temperature on fly ball distance somewhere on the web, might have been SoSh or Baseball Fever… that combined with the work Tango did recently on the relationship between fly ball distance and home run rates should add to this discussion.  I’ll find it later if I can’t locate it on the web…

I do have this: here is a link to a .jpg showing average game time temperature by month and by park.  It should be obvious that month and ballpark, and thus roster, are confounded, and will need some careful detangling before the data can be interpreted with much confidence.

Note especially the cold Pacific Coast parks: I forgot to mention San Diego in my earlier post - average temps under 70 for 1st half of season, then barely 70 the 2nd half…

http://www.hittrackeronline.com/AvgTemps.jpg